Examining the Pro-Eating Disorders Community on Twitter Via the Hashtag #proana: Statistical Modeling Approach

Page created by Victoria Chen
 
CONTINUE READING
JMIR MENTAL HEALTH                                                                                                         Sukunesan et al

     Original Paper

     Examining the Pro-Eating Disorders Community on Twitter Via
     the Hashtag #proana: Statistical Modeling Approach

     Suku Sukunesan1, PhD; Minh Huynh2, PhD; Gemma Sharp3, PhD
     1
      Information Systems Deptartment, Swinburne University of Technology, Hawthorn, Australia
     2
      Department of Dietetics, La Trobe University, Melbourne, Australia
     3
      Monash Alfred Psychiatry Research Centre, Monash University, Melbourne, Australia

     Corresponding Author:
     Suku Sukunesan, PhD
     Information Systems Deptartment
     Swinburne University of Technology
     John St
     Hawthorn, 3122
     Australia
     Phone: 61 9213 4373
     Email: ssinnappan@swin.edu.au

     Abstract
     Background: There is increasing concern around communities that promote eating disorders (Pro-ED) on social media sites
     through messages and images that encourage dangerous weight control behaviors. These communities share group identity formed
     through interactions between members and can involve the exchange of “tips,” restrictive dieting plans, extreme exercise plans,
     and motivating imagery of thin bodies. Unlike Instagram, Facebook, or Tumblr, the absence of adequate policy to moderate
     Pro-ED content on Twitter presents a unique space for the Pro-ED community to freely communicate. While recent research has
     identified terms, themes, and common lexicon used within the Pro-ED online community, very few have been longitudinal. It is
     important to focus upon the engagement of Pro-ED online communities over time to further understand how members interact
     and stay connected, which is currently lacking.
     Objective: The purpose of this study was to explore beyond the common messages of Pro-ED on Twitter to understand how
     Pro-ED communities get traction over time by using the hashtag considered to symbolize the Pro-ED movement, #proana. Our
     focus was to collect longitudinal data to gain a further understanding of the engagement of Pro-ED communities on Twitter.
     Methods: Descriptive statistics were used to identify the preferred tweeting style of Twitter users (either as mentioning another
     user in a tweet or without) as well as their most frequently used hashtag, in addition to #proana. A series of Mann Whitney U
     tests were then conducted to compare preferred posting style across number of followed, followers, tweets, and favorites. This
     was followed by linear models using a forward step-wise approach that were applied for Pro-ED Twitter users to examine the
     factors associated with their number of followers.
     Results: This study reviewed 11,620 Pro-ED Twitter accounts that posted using the hashtag #proana between September 2015
     and July 2018. These profiles then underwent a 2-step screening of inclusion and exclusion criteria to reach the final sample of
     967 profiles. Over 90% (10,484/11,620) of the profiles were found to have less than 6 tweets within the 34-month period. Most
     of the users were identified as preferring a mentioning style of tweeting (718/967, 74.3%) over not mentioning (248/967, 25.7%).
     Further, #proana and #thinspo were used interchangeably to propagate shared themes, and there was a reciprocal effect between
     followers and the followed.
     Conclusions: Our analysis showed that the number of accounts followed and number of Pro-ED tweets posted were significant
     predictors for the number of followers a user has, compared to likes. Our results could potentially be useful to social media
     platforms to understand which features could help or otherwise curtail the spread of ED messages and activity. Our findings also
     show that Pro-ED communities are transient in nature, engaging in superficial discussion threads but resilient, emulating
     cybersectarian behavior.

     (JMIR Ment Health 2021;8(7):e24340) doi: 10.2196/24340

     https://mental.jmir.org/2021/7/e24340                                                       JMIR Ment Health 2021 | vol. 8 | iss. 7 | e24340 | p. 1
                                                                                                            (page number not for citation purposes)
XSL• FO
RenderX
JMIR MENTAL HEALTH                                                                                                             Sukunesan et al

     KEYWORDS
     Twitter; infodemiology; eating disorders; proana; thinspo; hashtags; transient; cybersectarianism

                                                                           images, videos, or web links, and users can contribute to larger
     Introduction                                                          conversations by adding keywords or hashtags within the tweet
     The prevalence of eating disorders (EDs) has been on the rise         [10,16,17].
     ever since the condition was listed in the Global Burden of           Hashtags can connect users and be used to form communities
     Disease Study [1]. Recent estimates show EDs claim the lives          around common interest topics [18]. In the online Pro-ED
     of 3.3 million people globally every year [2], a number that has      community, #proana signifies a post supporting pro-ED attitudes
     doubled over the last 10 years [3]. Of all the ED types, anorexia     and behaviors and is considered to be the established term to
     nervosa (AN) in particular poses severe life-threatening health       describe the Pro-ED movement’s consistent referencing of EDs
     risks, with the highest mortality rate of all mental illnesses [4].   (eg, explicit mentions of bulimia and AN) within these accounts.
     In addition, nonfatal presentations are listed as the fifth cause     Most of these accounts have acquired followers who themselves
     of chronic disease among adolescents aged 15-19 years old in          posted about EDs [10]. Other research has focused on
     the Australian female population [5].                                 #thinspiration (“motivating” imagery of thin bodies) and
     Traditional media platforms and their representation of the “thin     #fitspiration (“motivating” imagery of “fit” bodies) and their
     ideal” have long been associated with body dissatisfaction, a         use within a variety of social media formats. Across social media
     known risk and maintenance factor of EDs [6,7]. However, the          platforms, typically #thinspiration encourages more weight loss
     media landscape has changed dramatically in recent years, and         behaviors with a stronger connected community than that of
     the last decade has seen a surge in social media use globally,        #fitspiration [17]. However, both have been found to essentially
     where recent figures show more than half of the world’s               share the same themes of encouraging guilt, dieting, and restraint
     population, or 3.8 billion people, are active on social media [8].    [19-21]. Nevertheless, longitudinal hashtag research within the
     The use of the internet to communicate using common online            Pro-ED communities is still limited. Since Twitter does not
     platforms has become more popular due to the increasing focus         currently have a policy for blocking such hashtags, unlike other
     on usability, the decreasing cost in access, and the ability of       social media sites such as Instagram, Facebook [22], and Tumblr
     communications to cross large geographical distances [9]. This        [23], it presents a unique space to use these freely and has played
     transition has seen the emergence of social and interpersonal         host to a shift to a space in which the Pro-ED community now
     support networks for users and in particular, the emergence of        communicates [10,16].
     pro-eating disorder (Pro-ED) communities online [10-12].              Furthermore, research to date has typically focused on the
     Pro-ED communities are a controversial subculture that                characteristics of the specific social networking sites for
     promotes positive attitudes toward EDs, namely AN                     interaction and overlooked the exploration of the broader
     (pro-anorexia/proana) and bulimia nervosa (pro-bulimia/promia).       implications of online communities. Indeed, a meta-analysis of
     These communities share content to promote thinness, provide          pro-anorexia and pro-bulimia website studies reported that the
     advice to other members, and glorify low body weight as ideal         main body of research has neglected the investigation of
     [13]. A shared group identity is formed through interactions          individual members that comprise the communities, including
     between community members and can involve the exchange of             their behaviors, motivations, and state of health, instead
     “tips,” restrictive dieting plans, extreme exercise plans, and        examining the role and content of the websites in community
     motivating imagery of thin bodies, also known as “thinspiration”      building. However, research suggests that the effects of personal
     or “thinspo” [11,14]. Boero and Pascoe [15] described these           social groups and peer behavior are prominent features in this
     communities as being able to “bring people together who rarely        space. Allison et al [24] proposed that the forces of social
     talk about their disorder face to face in non-therapeutic settings”   imitation and competition drive group behavior and put forward
     and noted that these groups are present online at their own will      the idea that the “authoritative voice” of AN partly results from
     with no formal offline equivalent.                                    the expectations of the social group. This finding was further
                                                                           supported by Ferguson et al [7], who suggested peer competition
     There is now a significant body of literature highlighting the        as more prominent than traditional media effects when looking
     way in which Pro-ED communities exist on the internet and in          at body dissatisfaction in teenage girls.
     particular on social networking platforms such as Twitter
     [10,16,17]. The Twitter platform is a social networking service       With this in mind, we suggest that focus needs to be placed
     known for its microblogging capability and is used by 339.6           upon the engagement of Pro-ED online communities to further
     million people, mostly between the ages 18 years and 34 years         understand how members interact, as Girvan and Newman [25]
     [8]. Twitter users can create a profile known as a “handle” and       demonstrated through the creation of social graphs in which
     post microblogs or “tweets,” typically text comprising 140            communication is visualized as relationships between entities.
     characters or less (although this was increased to 280 characters     In addition, recent research suggests identifying terms, themes,
     in 2017) from which other users can then share, known as a            and a common lexicon used within the Pro-ED online
     “retweet,” or follow other users to create their own personal,        community as beneficial in understanding a Pro-ED identity
     interconnected social network. The platform has become a center       [10,16]. Choudhury [26] looked at Tumblr to understand how
     for online social activity and the quick exchange of information,     both pro-anorexia and pro-recovery communities interact
     with the option now to post content other than text such as           through tags and a common lexicon, with findings suggesting

     https://mental.jmir.org/2021/7/e24340                                                           JMIR Ment Health 2021 | vol. 8 | iss. 7 | e24340 | p. 2
                                                                                                                (page number not for citation purposes)
XSL• FO
RenderX
JMIR MENTAL HEALTH                                                                                                            Sukunesan et al

     that AN content can be detected with a high level of accuracy
     due to distinctive “affective, social, cognitive, and linguistic
                                                                          Methods
     style markers.” Chancellor et al [14] further explored the lexicon   Ethics Approval
     of Pro-ED community members, this time on Instagram, both
     before and after attempts of moderation by the social networking     The current study was approved by the Swinburne University
     site to create a codebook of variations used to circumvent           Human Research Ethics Committee (SUHREC) Project ID:
     restrictions. A similar codebook of keywords was developed           20190402-1922. In line with SUHREC advice, it was not
     by Arseniev-Koehler et al [10] and Zhou et al [16] for Twitter,      possible to directly quote individual twitter usernames or their
     in an attempt to summarize and describe ED content. However,         posts; thus, data are presented in aggregated form only.
     both of the studies did not track user profiles over time, and in    Sample
     particular, their study did not consider the tweeting styles of
                                                                          This study utilized publicly available Twitter data from Pro-ED
     community members that could provide insights into how
                                                                          profiles collected between September 15, 2015 and July 1, 2018,
     Pro-ED communities communicate and interact.
                                                                          adhering to university ethics requirements. To identify Pro-ED
     Our study sought to extend upon previous examinations of             profiles, we used an online scraping tool to gather posts (tweets)
     Pro-ED tweets and in particular examine profiles and                 and reposts (retweets) using Twitter's public Application
     engagement of Pro-ED communities together with preferred             Program Interface (API). Twitter offers a systematic collection
     tweeting styles among Pro-ED users.                                  of sampled tweets as they are posted through a public API
                                                                          filtered by specific criteria. For this research, the hashtag
     A secondary objective was to identify the most frequently used
                                                                          #proana was the qualifying criteria, which resulted in 54,506
     hashtag among Twitter profiles that include “proana” as a
                                                                          tweets and retweets (tweets that are recirculated by other users)
     primary hashtag. A greater understanding of the Pro-ED
                                                                          across 11,620 Twitter profiles from various time zones and
     communication networks on Twitter could have implications
                                                                          geographic locations. These profiles then underwent a double
     for the identification, prevention, and treatment of young people
                                                                          pseudonymization process to preserve anonymity before a 2-step
     with EDs who may be receptive to online therapeutic
                                                                          screening process using inclusion and exclusion criteria was
     interventions.
                                                                          imposed to reach the final sample of 967 profiles (see Figure
                                                                          1).
     Figure 1. Selection of the sample for this study.

     https://mental.jmir.org/2021/7/e24340                                                          JMIR Ment Health 2021 | vol. 8 | iss. 7 | e24340 | p. 3
                                                                                                               (page number not for citation purposes)
XSL• FO
RenderX
JMIR MENTAL HEALTH                                                                                                                            Sukunesan et al

     Data Analysis                                                                   Skewed predictor variables were trimmed by excluding extreme
     The 967 profiles were further classified into 1 of 2 categories                 cases, as identified with a Cook distance >3 SDs from the mean.
     (with-mention or without-mention) based upon the user’s                         Analyses were conducted in SPSS Version 26.0 (IBM Corp,
     preferred posting style. A without-mention message pertained                    Armonk, NY).
     to the user sending a tweet not to a particular individual, whereas
     a mention relates to the user including another Twitter account                 Results
     in the message. In the initial phase, descriptive statistics were
                                                                                     The individual tweets or retweets across all 967 profiles were
     ascertained to compare the proportion of tweets or retweets in
                                                                                     explored to ascertain the most frequently used hashtag associated
     relation to the user’s most frequently used hashtag. A series of
                                                                                     with each account, as shown in Table 1. The “Top Hashtag”
     Mann Whitney U tests were then conducted to compare
                                                                                     variable represents the most frequently used hashtag among
     preferred posting style across number of followed, followers,
                                                                                     users; for example, for this sample, 611 users (63.2%) had
     tweets, and favorites. Finally, a multiple linear regression model,
                                                                                     #thinspo as their most frequently used hashtag (excluding
     using ordinary least squares [27], was then used to estimate the
                                                                                     #proana). In contrast, the “Hashtag usage” variable relates to
     number of followers based upon the number of followed users,
                                                                                     how many individual tweets across all accounts included the
     tweets, and favorites. The criteria for stepwise selection were
                                                                                     said hashtag.
     based upon changes in the adjusted R2 values at each new step.

     Table 1. Comparing the top 10 most used hashtags (excluding proana) in this sample (967 users and 54,506 tweets or retweets).
         Hashtag                                          Top hashtag, n (%)                                Hashtag usage, n (%)
         Thinspo                                          611 (63.2)                                        10,854 (20.53)
         41DaysofStarvation                               46 (4.8)                                          172 (0.33)
         Ana                                              39 (4.0)                                          2615 (4.95)
         Thinspiration                                    38 (4.9)                                          4499 (8.51)
         Promia                                           31 (3.2)                                          2549 (4.82)
         Anorexia                                         30 (3.1)                                          2537 (4.80)
         Redbraceletpro                                   29 (3.0)                                          845 (1.60)
         Skinny                                           21 (2.2)                                          2930 (5.54)
         Bonespo                                          18 (1.9)                                          2429 (4.59)
         ED book review                                   13 (1.3)                                          394 (0.75)
         Total                                            876 (90.6)a                                       29,824 (56.42)a

     a
         Only the top 10 hashtags are shown; thus, the % values do not equal 100.

     The hashtag #41DaysOfStarvation was the most used hashtag                       tweeting style. Overall, most profiles were classified as
     for 46 users (46/967, 4.8%); this was the second highest category               preferring “with-mention” tweeting styles (718/967, 74.3%)
     after #proana and #thinspo (Table 1). Conversely,                               over “without-mention” tweeting styles (248/967, 25.7%). Table
     #41DaysofStarvation was only mentioned in 172 (172/54,506,                      2 displays the descriptive statistics for the 2 groups across the
     0.33%) of the total tweets and retweets in this sample.                         number of (1) profiles followed, (2) followers, (3) tweets, and
                                                                                     (4) favorites. There were no significant differences between the
     The 967 profiles were further classified into either
                                                                                     2 groups for all of the categories (followed, followers, tweets,
     “without-mention” or “with-mention” groups based upon their
                                                                                     and favorites).

     Table 2. Online behaviors grouped by tweeting behavior.
         Behaviors                Without mention (n=248)                           With mention (n=718)                           Mann-Whitney U            P
                                                                                                                                   test Z score
                                  Mean (SD)             Median                      Mean (SD)              Median
         Followed                 778.55 (2269.93)      168.5                       758.23 (1938.82)       187.5                   –0.88                     .381
         Followers                1001.16 (3605.59)     235.0                       887.11 (2192.43)       210.0                   –0.42                     .675
         Tweets                   10390.46 (46922.71)   876.0                       9886.84 (26378.60)     1281.0                  –1.36                     .173
         Favorites                3476.15 (8781.05)     483.5                       4228.01 (11702.89)     570.5                   –0.74                     .462

     The relationships between the aforementioned factors (see Table
     2) were also examined via a Pearson correlation test (see Table

     https://mental.jmir.org/2021/7/e24340                                                                          JMIR Ment Health 2021 | vol. 8 | iss. 7 | e24340 | p. 4
                                                                                                                               (page number not for citation purposes)
XSL• FO
RenderX
JMIR MENTAL HEALTH                                                                                                                     Sukunesan et al

     3). The results indicated that all the factors were significantly             between number of followers and accounts followed.
     correlated with each other, with the largest correlation being

     Table 3. Intercorrelations of online behaviors (N=966).
         Variable                                   Followed                           Followers                        Tweets
         Followers
             r                                      0.85                               1                                —a
             P value
JMIR MENTAL HEALTH                                                                                                             Sukunesan et al

     to be more community driven. The type of tweet (with-mention          removing postings beyond the immediate network of an
     vs without-mention) did not differ significantly across followed,     individual where postings are transversed fluidly and randomly.
     followers, tweets, and favorites. This implies that whether a         Social media platforms will need to take heed of the fact that
     member is the source of the content or merely sharing it, they        user content lives beyond the immediate layer of where a posting
     are equally likely to contribute to the growing Pro-ED                has been initially lodged and could be shared across different
     community and its formation. As previously suggested by Wang          platforms. This could impact policy development for content
     et al [31], members within the Pro-ED Twitter community use           removal and moderation to avoid similar incidents to the live
     the platform as a tool for community engagement and not               streaming of a mass shooting in Christchurch, New Zealand via
     typically as a means of communication per se as indicated by          Facebook [36]. To best address these and other contextual issues,
     the number of retweets within our findings. This is a crucial         social media platforms need to work closely with external
     finding as there is a greater role that social media platforms can    support organizations to adopt a best practice approach. In the
     play in addressing the communication. In essence, social media        context of ED, Twitter will need to soon adopt national ED
     platforms could fill this void with tools that can facilitate         bodies as safety partners [37] to continuously engage and be
     communication and extend ED-related discussions with the ED           advised on matters relating to ED.
     community users. One approach would be to channel them to
                                                                           From our analysis, there appears to be a reciprocal effect
     external sites, such as the National Butterfly Foundation (official
                                                                           between followers and the followed. This implies that the
     organization for ED-related matters in Australia), mediated by
                                                                           Pro-ED community is resilient [38] and gains traction, as more
     a chatbot for cost efficiency.
                                                                           and more people may be influenced to be part of it. This
     The Pro-ED community chatter was dominated by retweets, by            emulates cybersectarian society behaviors [39], whereby niche
     75%, rather than genuine threads of communication. For                sentiments appeal to only a select community of people who
     example, the exchanges featuring “41DaysofStarvation” were            propagate information and are virtually enduring. While opinion
     a passing superficial topic in a particular subgroup of users that    leaders and influencers have been found to exist within online
     garnered quick interest and then discontinued. This could be          ED communities [31], dominating members are not typically
     due to the members’ transient [32-34] nature, which prohibits         apparent. Dominating members can exert constant enforcement
     them from building longer-lasting discussion threads, with over       or exhibit power that could encourage members to change their
     10,484 profiles only engaging in no more than 6 tweets. It is         allegiance behavior or even abandon the community [35]. As
     possible that this “41 days” of extreme weight control led to a       indicated through the type of tweets and number of retweets,
     deterioration in their physical health and subsequent inpatient       the Pro-ED community engages with content and propagates
     admission. However, this pattern of communication could also          it, and while externally, the community may appear just as an
     be indicative of the network structure. Previous findings indicate    avenue for individuals seeking social support, the focus is
     that Pro-ED communities have a far-reaching online community          potentially more about aligning with the collective identity of
     [15] but low reciprocity rates of communication with other users      the community. Both issues with identity and social roles have
     through replies and mentions [31]. This alludes to the allegiance     been noted as risk and maintenance factors of EDs [31]. Adverse
     the members have towards the topic and care shown among               health outcomes of these groups have been observed over time
     members but rarely are there extended discussion threads [35].        on social media platforms in their desire to become “thin,” hence
     Additionally, the possibility of users being barred for violating     the crucial need for an understanding of the community structure
     the rules of engagement, especially if their postings included        and development of innovative intervention methods. When
     suicidal and self-harm messages, may account for hashtag              faced with mediation, cybersectarian groups typically react
     attrition rates. This was evident within this study where 14.9%       impulsively to go incognito and reappear after a length of time
     (169/1136) of the Twitter users had their account suspended or        or remain hidden forever. For the health and safety of the
     deleted between September 15, 2015 and July 1, 2018 as shown          members of these groups, a more participatory treatment
     in Figure 1. A further analysis on the remaining 967 profiles         intervention would potentially generate better outcomes
     revealed that only 632 profiles are currently still active, showing   compared with an outright ban, as noted by Casilli et al [40].
     a 34% attrition from July 1, 2018 to September 15, 2020. For
                                                                           Understanding the vitality of Pro-ED communities is relatively
     example, one Twitter user was barred for 2 months by Twitter
                                                                           complex and is reliant on the emergence of health fads and the
     for posting adverse Pro-ED content. This resulted in the removal
                                                                           traction of passing themes. Here, social media platforms such
     of all previous postings and interactions. This account holder
                                                                           as Twitter would need to play a proactive role in addressing
     has since resumed being online with the same Twitter handle
                                                                           these issues. For example, Twitter should directly communicate
     continuing posting Pro-ED content, however less active. This
                                                                           to the 632 active profiles reported in this study to reduce further
     incident was documented by the authors due to the longitudinal
                                                                           ED-related discussion and minimize sharing of related content
     data collected from Twitter, a strength of our study.
                                                                           that has a negative impact, as reported by Tiggemann and
     The data also showed retweets of the original postings being          Zaccardo [41]. While Twitter has already undertaken some
     still “alive” on Twitter despite the corrective action by the         action within the suicide and self-harm space [42], more would
     platform. This leads us to question the amount of ED-related          be expected to follow, as Boyd [43] noted that adolescent users
     messages that could be retweeted and commented upon long              frequently turn to social media platforms including Twitter as
     after an account has been deleted. Impact of these unhealthy          a coping mechanism to diffuse external pressures threatening
     messages could be everlasting to the society. This warrants           their mental health. It would also be beneficial for this approach
     further investigation but also highlights the complexity in           of analysis to be replicated on other social media platforms to

     https://mental.jmir.org/2021/7/e24340                                                           JMIR Ment Health 2021 | vol. 8 | iss. 7 | e24340 | p. 6
                                                                                                                (page number not for citation purposes)
XSL• FO
RenderX
JMIR MENTAL HEALTH                                                                                                          Sukunesan et al

     observe whether Pro-ED communities behave in the same               Limitations
     manner across platforms. Importantly, future research should        There are several limitations to the current study. First, some
     address how hashtags and other message content can be utilized      data may have been omitted in the data collection process owing
     to identify and reach individuals who are struggling with EDs       to the free data access from the Twitter streaming API, which
     and provide them with much needed therapeutic interventions.        normally constitutes about 1% of the whole Twitter data stream
     However, a challenge for interventions is the rapidly changing      [48]. However, as mentioned by Cavazos-Regh et al [28], the
     lexicon of the community [44]. As our findings indicate,            percentage of private Twitter accounts is very small, and Twitter
     hashtags accompanying Pro-ED events such as                         accounts default to a public setting. Results might be more
     #41DaysofStarvation were short lived; however, #proana persists     reflective if we had subscribed to Twitter premium API services
     as a consistent theme in an otherwise transient community,          [49] and targeted tweets from personal accounts, and a larger
     potentially providing an ideal starting point for intervention.     sample size would have made this study more generalizable
     Our analysis showed that the number of accounts followed and        across the board. Further, a suite of other ED-related hashtags
     number of Pro-ED tweets posted were significant predictors for      described in [16] would have contributed to a larger data set.
     the number of followers of a user compared to likes. Hence the      These factors will be considered in future to improve research
     “like” counter is an obsolete predictor for ED engagement and       outcomes.
     activity. This important finding could potentially be useful to
                                                                         Conclusions
     social media platforms to understand which features could help
     or otherwise curtail the spread of ED messages. A recent report     Notwithstanding these limitations, our study contributes to the
     about Instagram’s decision to turn off the “like” counter [45]      emerging literature examining Pro-ED content on social media
     might be futile to curtail ED, though the number of likes has       platforms by providing an understanding of the Pro-ED
     been reported to give some indication of support [46,47].           communities and also the engagement of these groups.
                                                                         Continued research is needed to understand how we might use
                                                                         these messages and group dynamics to provide intervention and
                                                                         support to people with EDs in need.

     Conflicts of Interest
     None declared.

     References
     1.      Hoek H. Review of the worldwide epidemiology of eating disorders. Curr Opin Psychiatry 2016 Nov;29(6):336-339. [doi:
             10.1097/YCO.0000000000000282] [Medline: 27608181]
     2.      van Hoeken D, Hoek HW. Review of the burden of eating disorders: mortality, disability, costs, quality of life, and family
             burden. Curr Opin Psychiatry 2020 Nov;33(6):521-527 [FREE Full text] [doi: 10.1097/YCO.0000000000000641] [Medline:
             32796186]
     3.      Galmiche M, Déchelotte P, Lambert G, Tavolacci M. Prevalence of eating disorders over the 2000-2018 period: a systematic
             literature review. Am J Clin Nutr 2019 May 01;109(5):1402-1413 [FREE Full text] [doi: 10.1093/ajcn/nqy342] [Medline:
             31051507]
     4.      Arcelus J, Mitchell AJ, Wales J, Nielsen S. Mortality rates in patients with anorexia nervosa and other eating disorders. A
             meta-analysis of 36 studies. Arch Gen Psychiatry 2011 Jul 04;68(7):724-731. [doi: 10.1001/archgenpsychiatry.2011.74]
             [Medline: 21727255]
     5.      Mathews RRS, Hall WD, Vos T, Patton GC, Degenhardt L. What are the major drivers of prevalent disability burden in
             young Australians? Med J Aust 2011 Mar 07;194(5):232-235. [doi: 10.5694/j.1326-5377.2011.tb02951.x] [Medline:
             21381994]
     6.      Grabe S, Ward LM, Hyde JS. The role of the media in body image concerns among women: a meta-analysis of experimental
             and correlational studies. Psychol Bull 2008 May;134(3):460-476. [doi: 10.1037/0033-2909.134.3.460] [Medline: 18444705]
     7.      Ferguson CJ, Muñoz ME, Garza A, Galindo M. Concurrent and prospective analyses of peer, television and social media
             influences on body dissatisfaction, eating disorder symptoms and life satisfaction in adolescent girls. J Youth Adolesc 2014
             Jan;43(1):1-14. [doi: 10.1007/s10964-012-9898-9] [Medline: 23344652]
     8.      Kemp S. Digital 2020: 3.8 Billion People Use Social Media. We Are Social. 2020 Jan 30. URL: https://wearesocial.com/
             blog/2020/01/digital-2020-3-8-billion-people-use-social-media [accessed 2020-07-02]
     9.      Hills P, Argyle M. Uses of the Internet and their relationships with individual differences in personality. Computers in
             Human Behavior 2003 Jan;19(1):59-70. [doi: 10.1016/s0747-5632(02)00016-x]
     10.     Arseniev-Koehler A, Lee H, McCormick T, Moreno MA. #Proana: Pro-Eating Disorder Socialization on Twitter. J Adolesc
             Health 2016 Jun;58(6):659-664. [doi: 10.1016/j.jadohealth.2016.02.012] [Medline: 27080731]
     11.     Borzekowski DLG, Schenk S, Wilson JL, Peebles R. e-Ana and e-Mia: A content analysis of pro-eating disorder Web sites.
             Am J Public Health 2010 Aug;100(8):1526-1534. [doi: 10.2105/AJPH.2009.172700] [Medline: 20558807]
     12.     Holland G, Tiggemann M. A systematic review of the impact of the use of social networking sites on body image and
             disordered eating outcomes. Body Image 2016 Jun;17:100-110. [doi: 10.1016/j.bodyim.2016.02.008] [Medline: 26995158]

     https://mental.jmir.org/2021/7/e24340                                                        JMIR Ment Health 2021 | vol. 8 | iss. 7 | e24340 | p. 7
                                                                                                             (page number not for citation purposes)
XSL• FO
RenderX
JMIR MENTAL HEALTH                                                                                                         Sukunesan et al

     13.     Bell V. Online information, extreme communities and internet therapy: Is the internet good for our mental health? Journal
             of Mental Health 2009 Jul 06;16(4):445-457. [doi: 10.1080/09638230701482378]
     14.     Chancellor S, Pater J, Clear T, Gilbert E, De Choudhury M. #thyghgapp: Instagram Content Moderation and Lexical
             Variation in Pro-Eating Disorder Communities. 2016 Presented at: ACM Conference on Computer-Supported Cooperative
             Work & Social Computing; February 27 - March 2, 2016; San Francisco, CA. [doi: 10.1145/2818048.2819963]
     15.     Boero N, Pascoe C. Pro-anorexia Communities and Online Interaction: Bringing the Pro-ana Body Online. Body & Society
             2012 May 24;18(2):27-57. [doi: 10.1177/1357034X12440827]
     16.     Zhou S, Bian J, Zhao Y, Haynos A, Rizvi R, Zhang R. Analysis of Twitter to Identify Topics Related to Eating Disorder
             Symptoms. IEEE Int Conf Healthc Inform 2019 Jun;2019:10 [FREE Full text] [doi: 10.1109/ichi.2019.8904863] [Medline:
             32030368]
     17.     Tiggemann M, Churches O, Mitchell L, Brown Z. Tweeting weight loss: A comparison of #thinspiration and #fitspiration
             communities on Twitter. Body Image 2018 Jun;25:133-138. [doi: 10.1016/j.bodyim.2018.03.002] [Medline: 29567619]
     18.     Java A, Song X, Finin T, Tseng B. Why We Twitter: An Analysis of a Microblogging Community. In: Zhang H H,
             Spiliopoulou M, Mobasher B, Giles CL, McCallum A, Nasraoui O, et al, editors. Advances in Web Mining and Web Usage
             Analysis. SNAKDD 2007. Lecture Notes in Computer Science, vol 5439. Berlin, Germany: Springer Publishing Company;
             2009:118-138.
     19.     Boepple L, Thompson JK. A content analytic comparison of fitspiration and thinspiration websites. Int J Eat Disord 2016
             Jan 16;49(1):98-101. [doi: 10.1002/eat.22403] [Medline: 25778714]
     20.     Casilli AA, Tubaro P, Araya P. Ten years of Ana: Lessons from a transdisciplinary body of literature on online pro-eating
             disorder websites. Social Science Information 2012 Mar 15;51(1):120-139. [doi: 10.1177/0539018411425880]
     21.     Cobb G. “This is pro-ana”: Denial and disguise in pro-anorexia online spaces. Fat Studies 2016 Dec 02;6(2):189-205. [doi:
             10.1080/21604851.2017.1244801]
     22.     Ostroff N, Taylor J. Tumblr to ban self-harm and eating disorder blogs. BBC News. 2012 Mar 26. URL: http://www.
             bbc.co.uk/newsbeat/17195865 [accessed 2020-09-08]
     23.     About Eating Disorders. Instagram. URL: https://help.instagram.com/252214974954612 [accessed 2020-06-08]
     24.     Allison S, Warin M, Bastiampillai T. Anorexia nervosa and social contagion: clinical implications. Aust N Z J Psychiatry
             2014 Mar 22;48(2):116-120. [doi: 10.1177/0004867413502092] [Medline: 23969627]
     25.     Girvan M, Newman MEJ. Community structure in social and biological networks. Proc Natl Acad Sci U S A 2002 Jun
             11;99(12):7821-7826 [FREE Full text] [doi: 10.1073/pnas.122653799] [Medline: 12060727]
     26.     De Choudhury M. Anorexia on Tumblr: A Characterization Study. 2015 Presented at: DH '15: 5th International Conference
             on Digital Health 2015; May 18-20, 2015; Florence, Italy. [doi: 10.1145/2750511.2750515]
     27.     Weisberg S. Applied linear regression. New Jersey: John Wiley and Sons; 2014.
     28.     Cavazos-Rehg PA, Krauss MJ, Costello SJ, Kaiser N, Cahn ES, Fitzsimmons-Craft EE, et al. "I just want to be skinny.":
             A content analysis of tweets expressing eating disorder symptoms. PLoS One 2019 Jan 16;14(1):e0207506 [FREE Full
             text] [doi: 10.1371/journal.pone.0207506] [Medline: 30650072]
     29.     Golder SA, Donath J. Social roles in electronic communities. 2004 Presented at: Association of Internet Researchers (AoIR)
             conference; September 2004; Brighton, UK. [doi: 10.4135/9781452244723.n29]
     30.     Angeletou S, Rowe M, Alani H. Modelling and Analysis of User Behaviour in Online Communities. In: Aroyo L, Welty
             C, Alani H, editors. he Semantic Web – ISWC 2011. ISWC 2011. Lecture Notes in Computer Science, vol 7031. Berlin,
             Germany: Springer Publishing Company; 2011:35-50.
     31.     Wang T, Brede M, Ianni A, Mentzakis E. Social interactions in online eating disorder communities: A network perspective.
             PLoS One 2018 Jul 30;13(7):e0200800 [FREE Full text] [doi: 10.1371/journal.pone.0200800] [Medline: 30059512]
     32.     Bond E. Virtually anorexic — Where's the harm? A research study on the risks of pro-anorexia websites. Children's Media
             Foundation. 2012 Nov 28. URL: https://www.thechildrensmediafoundation.org/wp-content/uploads/2014/02/
             Bond-2012-Research-on-pro-anorexia-websites.pdf [accessed 2021-06-29]
     33.     Griffiths F, Dobermann T, Cave J, Thorogood M, Johnson S, Salamatian K, et al. The Impact of Online Social Networks
             on Health and Health Systems: A Scoping Review and Case Studies. Policy Internet 2015 Dec;7(4):473-496 [FREE Full
             text] [doi: 10.1002/poi3.97] [Medline: 27134699]
     34.     Lehmann J, Castillo C, Lalmas M, Zuckerman E. Transient News Crowds in Social Media. 2013 Presented at: Seventh
             International AAAI Conference on Weblogs and Social Media; July 2013; Boston, MA URL: https://www.aaai.org/ocs/
             index.php/ICWSM/ICWSM13/paper/viewFile/6107/6374
     35.     Preece J. Sociability and usability in online communities: Determining and measuring success. Behaviour & Information
             Technology 2010 Nov 08;20(5):347-356. [doi: 10.1080/01449290110084683]
     36.     Quodling A. Anxieties over livestreams can help us design better Facebook and YouTube content moderation. The
             Conversation. URL: https://theconversation.com/
             anxieties-over-livestreams-can-help-us-design-better-facebook-and-youtube-content-moderation-113750 [accessed
             2020-09-01]
     37.     Trust and Safety Council. Twitter. URL: https://about.twitter.com/en_us/safety/safety-partners.html [accessed 2021-06-29]

     https://mental.jmir.org/2021/7/e24340                                                       JMIR Ment Health 2021 | vol. 8 | iss. 7 | e24340 | p. 8
                                                                                                            (page number not for citation purposes)
XSL• FO
RenderX
JMIR MENTAL HEALTH                                                                                                                  Sukunesan et al

     38.     Garcia D, Mavrodiev P, Schweitzer F. Social Resilience in Online Communities: The Autopsy of Friendster. 2013 Presented
             at: First ACM conference on online social networks; October 2013; Boston, MA p. 07-08. [doi: 10.1145/2512938.2512946]
     39.     Thornton P. The New Cybersects: Resistance and Repression in the Reform era. In: Perry E, Selden M, editors. Chinese
             Society: Change, Conflict and Resistance. London, England: Routledge; 2003:149-150.
     40.     No authors. Online networks of eating-disorder websites: why censoring pro-ana might be a bad idea. Perspect Public
             Health 2013 Mar;133(2):94-95. [doi: 10.1177/1757913913475756] [Medline: 23467530]
     41.     Tiggemann M, Zaccardo M. "Exercise to be fit, not skinny": The effect of fitspiration imagery on women's body image.
             Body Image 2015 Sep;15:61-67. [doi: 10.1016/j.bodyim.2015.06.003] [Medline: 26176993]
     42.     What to do about self-harm and suicide concerns on Twitter. Twitter. URL: https://help.twitter.com/en/safety-and-security/
             self-harm-and-suicide [accessed 2021-06-29]
     43.     Boyd D. It's complicated: the social lives of networked teens. New Haven, CT: Yale University Press; 2014.
     44.     Gerrard Y. Beyond the hashtag: Circumventing content moderation on social media. New Media & Society 2018 May
             28;20(12):4492-4511. [doi: 10.1177/1461444818776611]
     45.     Faucher K. Is Instagram's removal of its "like" counter a turning point in social media? The Conversation. 2019 Jul 21.
             URL: https://theconversation.com/is-instagrams-removal-of-its-like-counter-a-turning-point-in-social-media-119064
             [accessed 2020-09-01]
     46.     Tiggemann M, Hayden S, Brown Z, Veldhuis J. The effect of Instagram "likes" on women's social comparison and body
             dissatisfaction. Body Image 2018 Sep;26:90-97. [doi: 10.1016/j.bodyim.2018.07.002] [Medline: 30036748]
     47.     Li P, Chang L, Chua THH, Loh RSM. “Likes” as KPI: An examination of teenage girls’ perspective on peer feedback on
             Instagram and its influence on coping response. Telematics and Informatics 2018 Oct;35(7):1994-2005 [FREE Full text]
             [doi: 10.1016/j.tele.2018.07.003]
     48.     Sampled Stream. Twitter Developer Platform. URL: https://developer.twitter.com/en/docs/twitter-api/tweets/sampled-stream/
             integrate/consuming-sampled-stream-data [accessed 2021-06-29]
     49.     Premium APIs. Twitter Developer Platform. URL: https://developer.twitter.com/en/products/twitter-api/premium-apis
             [accessed 2021-06-29]

     Abbreviations
               AN: anorexia nervosa
               API: application program interface
               ED: eating disorder
               Pro-ED: promote eating disorders
               SUHREC: Swinburne University Human Research Ethics Committee

               Edited by J Torous; submitted 15.09.20; peer-reviewed by S Oska, E van Furth, M Roncero; comments to author 01.12.20; revised
               version received 09.12.20; accepted 25.05.21; published 09.07.21
               Please cite as:
               Sukunesan S, Huynh M, Sharp G
               Examining the Pro-Eating Disorders Community on Twitter Via the Hashtag #proana: Statistical Modeling Approach
               JMIR Ment Health 2021;8(7):e24340
               URL: https://mental.jmir.org/2021/7/e24340
               doi: 10.2196/24340
               PMID: 34255707

     ©Suku Sukunesan, Minh Huynh, Gemma Sharp. Originally published in JMIR Mental Health (https://mental.jmir.org), 09.07.2021.
     This is an open-access article distributed under the terms of the Creative Commons Attribution License
     (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium,
     provided the original work, first published in JMIR Mental Health, is properly cited. The complete bibliographic information, a
     link to the original publication on https://mental.jmir.org/, as well as this copyright and license information must be included.

     https://mental.jmir.org/2021/7/e24340                                                                JMIR Ment Health 2021 | vol. 8 | iss. 7 | e24340 | p. 9
                                                                                                                     (page number not for citation purposes)
XSL• FO
RenderX
You can also read